You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@mesos.apache.org by "Justin Clayton (JIRA)" <ji...@apache.org> on 2016/05/04 17:19:12 UTC
[jira] [Commented] (MESOS-3532) 3 Master HA setup restarts every 3
minutes
[ https://issues.apache.org/jira/browse/MESOS-3532?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15271020#comment-15271020 ]
Justin Clayton commented on MESOS-3532:
---------------------------------------
We're seeing this exact same issue. The only thing we've been able to do is switch to a single master for now. Any workarounds to this issue would be appreciated.
> 3 Master HA setup restarts every 3 minutes
> ------------------------------------------
>
> Key: MESOS-3532
> URL: https://issues.apache.org/jira/browse/MESOS-3532
> Project: Mesos
> Issue Type: Bug
> Affects Versions: 0.23.0
> Reporter: Edward Donahue III
> Attachments: master01.tar, master02.tar, master03.tar
>
>
> CentOS 7.1, 3 Node cluster, each host has mesos master/slave and zookeeper setup.
> After I pushed out a bad zoo.cfg (added 2 extra zookeeper hosts that didn't exist) about every three minutes the elected master restarts and this keeps happening, when I have just one of the three masters running, it restarts every 3 minutes.
> I fixed the configs, deleted all the files under (/var/log/zookeeper/version-2/, /var/lib/zookeeper/version-2/). Is there another step I need to take, I feel like zookeeper is the issue (also where I lack knowledge), this cluster was stable for months until I push out the bad zoo.cfg.
> The master logs have this output every second:
> I0928 13:56:05.281518 28448 replica.cpp:641] Replica in EMPTY status received a broadcasted recover request
> I0928 13:56:05.351608 28450 replica.cpp:641] Replica in EMPTY status received a broadcasted recover request
> I0928 13:56:05.351794 28448 recover.cpp:195] Received a recover response from a replica in EMPTY status
> I0928 13:56:05.352700 28452 recover.cpp:195] Received a recover response from a replica in EMPTY status
> I0928 13:56:05.352963 28447 recover.cpp:195] Received a recover response from a replica in VOTING status
> The mesos-slaves don't even register in time:
> I0928 13:55:40.041491 28418 slave.cpp:3087] master@10.251.132.179:5050 exited
> W0928 13:55:40.041574 28418 slave.cpp:3090] Master disconnected! Waiting for a new master to be elected
> E0928 13:55:40.250059 28420 socket.hpp:107] Shutdown failed on fd=9: Transport endpoint is not connected [107]
> I0928 13:55:48.005607 28418 detector.cpp:138] Detected a new leader: (id='14')
> I0928 13:55:48.005836 28417 group.cpp:656] Trying to get '/mesos/info_0000000014' in ZooKeeper
> W0928 13:55:48.006597 28417 detector.cpp:444] Leading master master@10.251.132.177:5050 is using a Protobuf binary f...ESOS-2340)
> I0928 13:55:48.006652 28417 detector.cpp:481] A new leading master (UPID=master@10.251.132.177:5050) is detected
> I0928 13:55:48.006731 28417 slave.cpp:684] New master detected at master@10.251.132.177:5050
> I0928 13:55:48.006891 28417 slave.cpp:709] No credentials provided. Attempting to register without authentication
> I0928 13:55:48.006911 28417 slave.cpp:720] Detecting new master
> I0928 13:55:48.006940 28417 status_update_manager.cpp:176] Pausing sending status updates
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)